Tesla and SpaceX Commit $16.8B to Texas 'Terafab' Chip Plant

The number worth staring at in Musk's new chip project isn't $16.8 billion, and it isn't even the $119 billion phased ceiling floating in a May SpaceX filing. It's 100 million square feet of manufacturing space under one roof, which is what Tesla and SpaceX are pitching for a jointly built fab in Grimes County, Texas.

According to TechCrunch, the two companies will initially spend $16.8 billion on 'Terafab,' a vertically integrated complex just north of Houston that will handle manufacturing, packaging, and testing of advanced logic and memory devices in one place. The chips are meant to feed Tesla's Optimus robots and Cybercabs, plus high-performance silicon for SpaceX's proposed space-based data centers. Elon Musk called it 'the largest and most valuable building on Earth by far,' and said Terafab will 'produce AI chips at scale for use on Earth and in space.' The site is expected to employ at least 3,000 people drawn from Grimes and neighboring Brazos counties.

Why this matters if you don't build robots or rockets: Musk is trying to move his AI stack out from under the same handful of foundries that everyone else in AI depends on. Bringing Intel on as a partner, which SpaceX did earlier this year, is a bet that owning the silicon end-to-end is worth the capex, and it hands Intel a marquee customer to prove its contract-manufacturing pivot is real.

The honest caveats are heavy. The $119 billion number is a filing aspiration across multiple future phases, not a commitment, and a May SpaceX filing described the first phase itself as a $55 billion proposal, well above what was announced this week. The reporting doesn't spell out what process node Terafab will run, what Intel's exact role or economic stake looks like, or a hard construction timeline for that initial tranche. Local scrutiny is already surfacing: a Grimes County meeting raised concerns about tax incentives, transparency, and SpaceX's plan to draw industrial water from Gibbons Creek Reservoir rather than local groundwater.

Take the specifics as reported, not settled. But if even the first phase gets built at anything close to the pitched scale, the balance of power in edge-inference silicon for robotics and autonomy shifts, and rivals dependent on external foundries will need to notice.